Inspiration Prediction markets are often treated as real-time truth engines, but confidence can mask instability and hype. We were inspired to question whether crowd belief actually reflects insight—or simply amplifies momentum. The Hype Index was created to examine how beliefs behave over time, not just what the final odds say.
What it does The Hype Index analyzes prediction-market data to measure confidence, volatility, sentiment flips, and hype decay. It turns raw probabilities into interpretable metrics and publishes them as an interactive data app with a conversational AI layer, allowing users to explore where markets are insightful and where they are overconfident.
How we built it We built the project end-to-end in Hex, using SQL and Python notebooks to clean and analyze Kalshi data. We engineered time-series metrics, created Hex-native visualizations and controls, defined a semantic model for trusted metrics, and layered an AI Thread on top for natural-language exploration.
Challenges we ran into Prediction-market data is noisy and unevenly timed, making it difficult to separate real signal from short-term swings. Designing metrics that captured hype without oversimplifying behavior—and ensuring the app published error-free—required careful iteration.
Accomplishments that we're proud of Built a fully Hex-native experience combining analytics, semantic modeling, and AI Created original metrics that explain market behavior, not just outcomes Delivered an interactive, conversational data app that tells a clear story
What we learned
High confidence alone is not a reliable signal—stability and consistency matter just as much. We also learned how semantic models and AI chat can dramatically expand access to complex analysis.
What's next for The Hype Index
Next, we plan to expand to elections, sports, and financial markets, and validate predictions against real outcomes. Long-term, The Hype Index could evolve into a real-time hype-risk monitoring system to help decision-makers separate signal from noise early.
Log in or sign up for Devpost to join the conversation.